Fix for getting account by id

[maartenvanvliet]

Hey 👋

I added a way to override the method_name for an operation. These are not unique per schema apparently.
Did some cursory exploration on how other stripe sdk's deal with this:

• php - the Account.retrieve without an id is not implemented
• node - special cased, Account.retrieve handles the case where id is null.
• ruby - special cased implementation, overrides the parent class' implementation
• go - adds a getById method.

I've gone with the go way of fixing this and added retrieve_by_id. Two other resources had the same issue but naming was more straightforward there.

Also, the method_on: "collection" Stripe operations are now skipped. They were duplicates of the "service" operations and I think only used for code generation by Stripe itself.

fixes #777

[nkezhaya]

@maartenvanvliet Thanks for working on this! Maybe it's just me, but retrieve_by_id seems less appealing than retrieve.

Is there a benefit to having an Account.retrieve/0 function? It should just return an error, no?

[maartenvanvliet]

I agree that it's aesthetically less pleasing to have this one exception to how it's done the sdk and naming this function retrieve_by_id.
Other options I've considered were to name Account.retrieve/0 => Account.get/0 or Account.show/0, which are also not great imo. Because the go-lang implementation went this way, I followed suit.

The Account.retrieve/0 function will return the results for your own account, so it serves a purpose there. Though this information can also be retrieved with the Account.retrieve/1 function if you know the account id.

I'm not particularly attached to any function name and think something can be said for all of them. If there's consensus on the issue I'll push the changes.

[nkezhaya]

Oh right, my mistake!

I'm not a go expert by any means, but I suspect that GetByID might better follow go naming conventions. I could see an argument for get() and get(id) because users might be used to it from using Ecto and other libraries, but there'd be a small burden of changing all instances of retrieve/* to get/*.

Personally, I don't see anything wrong with Resource.retrieve() and Resource.retrieve(id).

[maartenvanvliet]

I mentioned that the two operations are not uniquely named in the schema but have failed to highlight the underlying issue. Thing is we cannot support retrieve/0-2 and retrieve/1-3 at the same time. The code generated leads to a compile issue where the defaults of the two functions clash. If this were possible it would be my preferred solution.

The simplest fix is to rename one of those functions to something else, that's why the overrides are there. The complicated fix is to have the generated code handle the different cases. This would imo hamper maintainability of the code for something that has 3 occurrences in the entire Stripe open api schema.

In light of this discussion, I propose to go with Account.show/0 for your own account, and keep Account.retrieve/1 to fetch another account.

[nkezhaya]

It's been a couple days and I'm unable to come up with anything better. 🙂 I like Account.show/0!

[maartenvanvliet]

@yordis I'll try to find some time to fix the conflicts and address the comments so this can be merged.

[yordis]

Sorry to be late to the party!

I am a bit confused. What is the difference between get_by_id/1 and retrieve/1?

Purely looking based on the test, Stripe.Account.retrieve("acct_123") should work, no?

I am used to Elixir leveraging pattern matching to avoid function explosion, so having a function that accepts a string or an object and things like that is expected.

[maartenvanvliet]

The issue is that the automatically generated functions clash in a few cases because of arity and defaults. In this case you'll get def retrieve/3 defaults conflicts with retrieve/2. That's why a method to override function names was introduced to bypass the clash.

[yordis]

CI failing, overall looking good 🚀 💜

[github-actions]

This pull request has been automatically marked as "stale:discard". If this pull request is still relevant, please leave any comment (for example, "bump"), and we'll keep it open. We are sorry that we haven't been able to prioritize reviewing it yet. Your contribution is very much appreciated!.

[gmrcp]

Hey 👋 do you have any updates on this?
I'm happy to help on the failing CI to push this forward

[yordis]

@gmrcp please do 🙏🏻

[maartenvanvliet]

CI failures are fixed, I also fixed an unrelated CI failure that was erroring with query params.

[yordis]

Should this be a breaking change? 🤔

[gjsduarte]

Hey everyone,
I'm really looking forward to this PR being merged.
Is there anything I can do to help you do so?

@maartenvanvliet, is there any chance you can resolve the conflicts?

[cgarvis]

@yordis anyway we can get this merged in?

[yordis]

@cgarvis I have a #853 where we must manually add ALL mapping. I do not want to make assumptions anymore; everything is explicit until we fix the anti-pattern (we should flatten the codegen and use operation ID as identity).

I would take any help with mix stripe.generate with the proper version (see makefile). It is a lot of grunt work!